This present invention relates generally to CMOS (complementary metal-oxide semiconductor) integrated circuits, and more particularly to I/O (input/output) structures and methods for such circuits.
Integrated circuits transmit and receive electrical signals to and from other circuitry using input and output xe2x80x9ccellsxe2x80x9d designed for that purpose. The physical connection between each input or output cell and outside circuitry is conventionally made by bonding a small wire to a bonding xe2x80x9cpadxe2x80x9d, i.e., an extended and exposed conductive region located on one of the circuit""s metal layers. For an input cell, receiving circuitry connects to the bonding pad. For an output cell, a transmitter or driver circuit connects to the bonding pad. Typically, both input and output cells also contain Electro-Static Discharge (ESD) protection circuitry that attempts to clamp large transient voltages (inadvertently applied to a bonding pad) before those voltages can damage a receiver or driver.
FIGS. 1, 2, and 3 illustrate three aspects of a simple output cell 20. Referring first to FIG. 1, P-channel MOS (PMOS) transistor 22 and N-channel (NMOS) transistor 24 operate as a complementary field-effect transistor (FET) pair signal driver. When signal IN is at a high voltage, transistor 22 is turned off and transistor 24 is turned on, pulling output pad 25 down towards Vss. Conversely, when signal IN is at a low voltage, transistor 24 is turned off and transistor 22 is turned on, pulling output pad 25 up towards Vdd.
PMOS transistor 26 and NMOS transistor 28 provide ESD protection for cell 20. Note that the gate of transistor 26 is permanently connected to Vdd, and the gate of transistor 28 is permanently connected to Vss, ensuring that these transistors are permanently off. But as shown in FIG. 2, transistors 26 and 28 contain diode structures that provide protection against voltage spikes. PMOS transistor 26 protects the cell from pad voltages much greater than Vdd, and NMOS transistor 28 protects the cell from pad voltages much less than Vss.
FIG. 3 shows a cross-section of transistors 26 and 28. Within PMOS transistor 26, a diode junction exists between the P+ drain diffusion 36 (connected to output pad 25) and the N-well drain diffusion 32 (connected to Vdd). Thus when the voltage at output pad 25 is slightly higher than Vdd, this diode junction is forward biased and current can flow from the pad to the Vdd voltage rail, clamping the pad voltage to a safe level.
Similarly, within NMOS transistor 28, a diode junction exists between the N+ drain diffusion 40 (connected to output pad 25) and the P-substrate 30 (connected to Vss). Thus when the voltage at output pad 25 is slightly lower than Vss, this diode junction is forward biased and current can flow from the Vss voltage rail to the pad, again clamping the pad voltage to a safe level.
Although transistors 26 and 28 in FIG. 1 are included for ESD protection, it is recognized herein that driver transistors 22 and 24 can have similarxe2x80x94albeit typically smaller due to smaller sizexe2x80x94ESD benefits if their bodies are biased appropriately. The described embodiments make use of this observation in an output cell having no (or reduced-size) ESD-only devices, augmented with multiple sets of driver transistors. The output cell contains a multimode logic circuit that, in each mode, configures at least some sets of driver transistors in an xe2x80x9coffxe2x80x9d mode that provides ESD protection.
For instance in one embodiment, an input/output cell connects to two pads. The cell has one set of differential drivers that allows a signal to be driven differentially on the two pads in one mode. The cell also has a set of single-ended drivers that allow two different signals to be driven on the two pads in another mode. In still another mode, the cell accepts input signals on the two pads. A multimode logic circuit selects the appropriate drivers for each mode, and turns off the remaining drivers, essentially placing them in an ESD mode.
One beneficial use of such an embodiment is in providing a flexible interface for an integrated circuit. Historically, a designer had to choose an interface type for each pad from a library of standard input and output cells. If two customers desired two different interface types, the circuit designer had to either design and manufacture two different integrated circuits, or provide two sets of pads and accompanying cells, one per interface type, on the circuit. Either approach is more expensive than the preferred embodiments described herein, which supply multiple different interface types on the same pads, at no significant increase in circuit area or cost.